In the present study we have demonstrated that extracellular S100A4 stimulates the expression of ephrin-A1 in NSCLC cell lines. Furthermore, we have characterized the expression of S100A4, ephrin-A1 and osteopontin in primary tumors from 217 NSCLC patients, and investigated the associations between these biomarkers and conventional clinicopathological parameters. Our group has previously shown that extracellular S100A4 induces the expression of ephrin-A1 and osteopontin in osteosarcoma cell lines by activating the transcription factor NF-κB [18
]. Based on these results, we wanted to investigate whether S100A4-mediated induction of ephrin-A1 and osteopontin also occurs in NSCLC cell lines. Interestingly, we observed that S100A4 was able to induce expression of ephrin-A1 both at the mRNA and protein level in adenocarcinoma, but not in squamous cell carcinoma cell lines. However, no S100A4-mediated stimulation of osteopontin expression was found in any of the cell lines tested. Importantly, a significant association was also found between expression of S100A4 and ephrin-A1 in primary tumor samples from NSCLC patients, indicating that S100A4 stimulates ephrin-A1 expression both in vivo and in vitro.
We found high expression of ephrin-A1 in 13% and intermediate expression in 72% of the tumors, and the fact that ephrin-A1 is expressed in the majority of the samples may suggest that this protein plays an important biological role in NSCLC. However, ephrin-A1 was not associated with any of the clinicopathological parameters apart from histological type. Interestingly, we found that adenocarcinomas had a higher percentage of S100A4 and ephrin-A1 positivity compared to squamous and large cell tumors, and this finding is in keeping with that of previous studies on S100A4 [7
] and ephrin-A1 [25
]. The histological subclasses of NSCLC differ not only in their presentation in different regions of the lung and in outcome [26
], but also in molecular characteristics and thereby in response to targeted therapies [27
]. Consequently, the differences in expression patterns of the protein markers between the adenocarcinomas and squamous cell carcinomas in this study are not surprising.
Expression of S100A4 in surgically resected NSCLC specimens has previously been investigated in several studies [7
], and the percentage of S100A4 positive cases in these studies range from 20-84%. In our study, intermediate or strong cytoplasmic expression of S100A4 was observed in 57% of the cases, which is comparable to the previous investigations. For osteopontin, high expression was found in 77% of the tumors, whereas in previous studies in NSCLC, osteopontin immunoreactivity range from 38–67% [20
]. In contrast to previous reports, where an association between high expression and squamous cell carcinoma has been described [28
], we did not find any significant associations between osteopontin expression and conventional clinicopathological parameters.
Possible explanations for the contradicting results for both S100A4 and osteopontin could be that different antibodies, different immunohistochemical staining techniques and different scoring systems were used. In the present study we have used immunohistochemical staining of tissue microarrays. A potential disadvantage with the use of TMA is the possibility that small tissue cores do not adequately represent the tumor, especially in cases with intratumoral heterogeneity. To evaluate whether the expression patterns of the protein markers on the small TMA cores were representative for the whole tumor, we immunostained seven whole sections with the same antibodies. The staining intensity of S100A4 and ephrin-A1 was generally homogenous across the sections, indicating that the obtained results are indeed representative of the whole tumor section. For osteopontin, however, some intratumor heterogeneity was observed. Also of importance, the majority of the mentioned studies have been retrospectively conducted, and the patient cohorts may therefore be biased. Our cohort was prospectively recruited, and the distribution of gender and age at surgery corresponds well with data from The Norwegian Association for Cardiothoracic Surgery. Thus, we believe that this patient population can be considered representative for patients with early stage NSCLC undergoing primary surgery in Norway.
S100A4 expression was associated with small tumor size and high degree of differentiation, and when analyzing the adenocarcinomas separately, significant inverse associations between S100A4 expression and lymph node metastasis as well as pTNM stage were found. Given that S100A4 in general is associated with poor prognosis and promotes metastasis in a number of tumor types [4
], this result was rather unexpected. Our results are also in contrast to other investigations in NSCLC where S100A4 expression was associated with high TNM stage and poor outcome [9
]. Importantly, in our cohort of prospectively recruited patients S100A4 expression was associated with several parameters that each reflects a less aggressive phenotype, suggesting that the observed result could be of clinical relevance, but further studies are required to clarify this issue.
How might we explain the unexpected result that S100A4 is associated with a non-aggressive phenotype in NSCLC? One of the most important biological functions contributing to S100A4-induced metastasis is increased cell migration and invasive capacity. However, induction of S100A4 has also been shown to decrease motility and invasiveness, such as in squamous cell carcinoma [31
], and down-regulation of S100A4 in astrocytes increased their migratory capacity in vitro [32
]. Furthermore, certain lines of evidence suggest that S100A4 may have tumor suppressor functions in the lung. S100A4 knockout mice, that were otherwise phenotypically normal, were prone to spontaneous tumor development, and the most frequent tumor observed was carcinoma of the lung [33
]. Taken together, these results indicate that the biological function of S100A4 is cell type-dependent, and possibly, S100A4 may not play a pro-metastatic role in all tumor types. One might also speculate that S100A4 could inhibit tumor progression in the early stages of NSCLC development, while promoting metastasis at later disease stages, similar to the cytokine transforming growth factor β [34
Moreover, our findings suggest that S100A4-induced expression of ephrin-A1 may be one mechanism by which S100A4 mediates its biological functions. If so, one should assume that similar functions are attributed to both proteins, and interestingly ephrin-A1 stimulates both cellular motility [15
], angiogenesis [13
] and metastasis [35
], features that are also associated with S100A4 [4
]. However, seemingly contradictory results have been reported for ephrin-A1, and overexpression of ephrin-A1 or treatment with ephrin-A1-Fc (soluble recombinant ephrin-A1 fused to the Fc portion of IgG) has been shown to inhibit invasiveness and reduce tumor growth in bladder, pancreatic and gastric cancer, and in malignant mesothelioma [36
]. In addition, ephrin-A1-Fc was found to inhibit tumor growth and migration in NSCLC cells [41
Ephrin-A1 is supposed to act as a tumor suppressor through its preferred receptor EphA2 [25
] which is overexpressed in NSCLC [41
Similar to its ligand, the role of EphA2 in cancer is somewhat conflicting. Increased expression is associated with poor clinical outcome in several tumor types, including NSCLC [3
]. However, EphA2 can also act as a tumor suppressor [43
], and recently, high expression of both EphA2 and ephrin-A1 was found to be related to favorable prognostic factors in stage I NSCLC patients [25
]. Based on our findings that S100A4 is associated with small tumor size and a less aggressive phenotype, one might speculate that S100A4-mediated induction of ephrin-A1 could be implicated in reduced tumor growth and invasiveness in NSCLC. However, ephrin-A1 expression was not associated with tumor size, differentiation or tumor stage, indicating that at least these S100A4-associated features are independent of ephrin-A1. Overall, these results suggest that ephrin-A1 plays an important role in tumor progression, but the exact function is complex, cell-type dependent and most likely relies on many factors, including its preferred receptor EphA2 [44
]. Furthermore, the role of ephrin-A1 as a biomarker still remains elusive, and especially in NSCLC further studies are certainly required.